On Mon, Oct 14, 2013 at 01:19:51PM +0200, Jérémie Courrèges-Anglas wrote: > Juan Francisco Cantero Hurtado <i...@juanfra.info> writes: > > > On Sun, Oct 13, 2013 at 10:03:23PM +0200, Jérémie Courrèges-Anglas wrote: > > [...] > > >> Would you mind trying this? This seems to fix the behavior of > >> mosh-server... > >> > >> I'm not asking for OKs since I don't understand what the issues really > >> are. If people with any C++ knowledge could explain what's wrong with > >> the original code in the second patch, thanks in advance. > > I sent this a bit too early, there is at least another "fix" needed for > the client (src/crypto/prng.h). Also... > > > The code compiles OK. mosh (i386 client, ARM server) doesn't work for me > > but maybe I've something broken on my ARM. > > > > MOSH CONNECT 0 u4id+P9EclW6TCZCbNDYXw > > Even if the server does bind to a port, Network::port() will always > return 0. Something is really fishy, I can't believe one could find > such an amount of brokenness in one port. I doubt it's just upstream > code. Even if we get around all this, I doubt it would be a reasonable > choice to let that stuff run...
Mosh is trying to bind to the port UDP 0. I see this output in amd64: MOSH CONNECT 60001 DLhBl0p7sQ9HI/Veu5i2ng -- Juan Francisco Cantero Hurtado http://juanfra.info
